[RP-PPPoE] Weird issues since moving to kernel-mode pppoe

[Extra Fu]

Hello,

Since I moved this morning to kernel-mode pppoe, the load  on the
server immediately disappeared, which is a very nice thing. I'm using
rp-pppoe 3.10 and pppd 2.4.4 on CentOS 5.3/x86.

It seems nonetheless that it has introduced two issues:

1- I keep seeing those messages in /var/log/messages since I activated
kernel-mode pppoe:

Feb 10 20:01:17 ppp1 pppd[14874]: Connect: ppp172 <--> eth1
Feb 10 20:01:17 ppp1 pppd[14874]: Couldn't increase MTU to 1500
Feb 10 20:01:17 ppp1 pppd[14874]: Couldn't increase MRU to 1500
Feb 10 20:01:19 ppp1 pppd[14874]: Couldn't increase MTU to 1500
Feb 10 20:01:19 ppp1 pppd[14874]: PAP peer authentication succeeded for foobar

while /etc/ppp/pppoe-server-options specifies:

mru 1492
mtu 1492

Any reasons?

2- Some users can establish their PPPoE connections but beside
ping'ing hosts (any on the Internet), nothing else works. DNS servers
are correctly pushed (since two valid entries are specified in
/etc/ppp/pppoe-server-options using ms-dns) and they can even be
ping'ed by users. For some (and most) other users, everything works as
expected.

I had none of those issues before moving to kernel-mode pppoe and
absolutely nothing has changed in the configuration (beside starting
pppoe-server using the -k parameter and loading rp-pppoe.so in
/etc/ppp/pppoe-server-options).

Any light that can be shed on those issues would be appreciated.

Best regards,